Effect of pleural invasion on survival of patients with small cell lung cancer: Propensity score analysis and nomogram establishment based on the SEER database.

OBJECTIVES

Pleural invasion (PI) is identified as an adverse prognostic factor for non-small cell lung cancer (NSCLC), but its value in small cell lung cancer (SCLC) remains unclear. We aimed to evaluate the survival effect of PI on overall survival (OS) in SCLC, meanwhile, we established a predictive nomogram based on related risk factors for OS in SCLC patients with PI.

METHODS

We extracted the data of patients diagnosed with primary SCLC between 2010 and 2018 from the Surveillance, Epidemiology, and End Results (SEER) database. The propensity score matching (PSM) method was used to minimize the baseline difference between the non-PI and PI groups. Kaplan-Meier curves and the log-rank test were used for survival analysis. Univariate and multivariate Cox regression analyses were applied to identify the independent prognostic factors. Randomly divided the patients with PI into training (70%) and validation (30%) cohorts. A prognostic nomogram was established based on the training cohort and was evaluated in the validation cohort. The C-index, receiver operating characteristic curves (ROC), calibration curves, and decision curve analysis (DCA) were applied to assess the performance of the nomogram.

RESULTS

A total of 1,770 primary SCLC patients were enrolled, including1321patients with non-PI and 449 patients with PI. After PSM, the 387 patients in the PI group matched the 387 patients in the non-PI group. By Kaplan-Meier survival analysis, we observed the exact beneficial effect of non-PI on OS in both original and matched cohorts. Multivariate Cox analysis showed similar results to demonstrate a statistically significant benefit for patients with non-PI in both original and matched cohorts. Age, N stage, M stage, surgery, radiotherapy, and chemotherapy were independent prognostic factors for SCLC patients with PI. The C-index of the nomogram in the training and validation cohort was 0.714 and 0.746, respectively. The ROC curves, calibration curves, and DCA curves also demonstrated good predictive performance in the training and validation cohorts of the prognostic nomogram.

CONCLUSION

Our study shows that PI is an independent poor prognostic factor for SCLC patients. The nomogram is a useful and reliable tool to predict the OS in SCLC patients with PI. The nomogram can provide strong references to clinicians to facilitate clinic decisions.